tao-train-mask-auto-encoder
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFE
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references official NVIDIA container images from the NVIDIA Container Registry (nvcr.io) for model training and deployment. These are established vendor resources necessary for the skill's stated purpose.\n- [COMMAND_EXECUTION]: Provides structured templates for executing model training and deployment tasks via the mae CLI and docker. These commands are scoped to the intended machine learning workflow.\n- [DATA_EXFILTRATION]: While the skill handles data ingestion from local or S3 sources, it does not contain hardcoded credentials or instructions to exfiltrate sensitive information. Parameters for encryption keys are provided as empty placeholders for user input.\n- [SAFE]: No malicious patterns, obfuscation, or safety bypass attempts were detected. The skill adheres to the NVIDIA TAO Toolkit's expected operational patterns.
Audit Metadata